Google has launched Gemini 3.7 Flash, a multimodal coding and agent-focused model with a 1M-token context window, 64K-token outputs, and March 2026 knowledge cutoff, delivered exclusively through APIs and enterprise platforms. The model refines Gemini 3.6 Flash with algorithmic improvements that boost software engineering, document-heavy knowledge work, and web development benchmarks, while keeping weights closed and excluding self-hosted or air-gapped deployments for organizations with strict data-residency requirements. Introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens undercuts Claude Sonnet 5 and GPT-5.6 Terra, encouraging startups and enterprises to test long-running agents and automation before rates double in 2027.
This update represents a notable development in the Ai sector. Organizations and founders tracking this space should evaluate potential strategic and technical implications on their operations.